DataFrame.interpolate method return type does not honor default inplace argument

[jasonboh]

Using pandas-stubs-2.0.1.230501.

inplace defaults to False, so the below shoudl return a new DataFrame. But pyright thinks the return type is DataFrame | None. Here is some code showing the problem:

import pandas as pd
df: pd.DataFrame = -pd.DataFrame()
df = df.interpolate(method="linear") # pyright reports the below message for this line

Expression of type "DataFrame | None" cannot be assigned to declared type "DataFrame"
  Type "DataFrame | None" cannot be assigned to type "DataFrame"
    Type "None" cannot be assigned to type "DataFrame"

But if I repalce the line with the below, then pyright is happy:

df = df.interpolate(method="linear", inplace=False)

[Dr-Irv]

Thanks for the report. I think the fix is to change

pandas-stubs/pandas-stubs/core/frame.pyi

Line 1711 in 002f0cc

inplace: Literal[False],

to be Literal[False] = ...

PR with test welcome.

[gandhis1]

FYI looks like this might also affect GroupBy.fillna: https://github.com/search?q=repo%3Apandas-dev%2Fpandas-stubs%20%22inplace%3A%20Literal%5BFalse%5D%2C%22&type=code

[Dr-Irv]

FYI looks like this might also affect GroupBy.fillna: https://github.com/search?q=repo%3Apandas-dev%2Fpandas-stubs%20%22inplace%3A%20Literal%5BFalse%5D%2C%22&type=code

Yes, I think you're correct. Thanks for finding that.

[jasonboh]

in fact, there are many, many places in the pandas API that use this same pattern, so if there's an easy way to check all the methods that have "inplace" args, that would be a good idea

[twoertwein]

FYI looks like this might also affect GroupBy.fillna: https://github.com/search?q=repo%3Apandas-dev%2Fpandas-stubs%20%22inplace%3A%20Literal%5BFalse%5D%2C%22&type=code

Docs say: Broken. Do not set to True. :)

[Dr-Irv]

Docs say: Broken. Do not set to True. :)

So we should fix the stubs to not allow the True value! Probably can remove the overloads and just change inplace to inplace: Literal[False] = ... and always have it return DataFrame